Oscillating Steam-Heater Loop: Valve Stiction or Aggressive Tuning?
A shell-and-tube steam heater supplies 82 °C water to a process. A single PID controller trims a 50 mm linear globe valve on 6 bar g saturated steam; the RTD in a 12 mm thermowell has a step response that reaches 63% in 22 s. Since a new controller was commissioned with PB = 15%, Ti = 0.8 min, Td = 0.05 min, the loop has had a sustained oscillation of about 90 s period and ±1.8 K amplitude. In manual, with the valve command held constant, the flow and temperature are steady. When the controller is put in manual and the output is stepped by 2%, the valve stem moves only after the command changes by about 1.5% and then jumps; stem position also drifts after the command stops. Plant wants to increase production, so the heater duty will rise by 15%.
1. Distinguish between valve stiction and over-aggressive PID tuning using the evidence given. Which is the more likely dominant cause?
Read the worked answer
The manual test is decisive. With the controller in manual and a constant output, temperature and flow are steady, so the process itself is not oscillating. A 2% output step that produces no stem movement until roughly 1.5% of command has accumulated, followed by a jump, is the signature of stick-slip: static friction holds the stem, the actuator builds force, and the valve then slips. The drift after the command stops indicates hysteresis or a loose linkage/positioner, not just high controller gain.
Purely aggressive tuning would also oscillate and would also stop in manual, so manual stability alone does not separate the two. The distinguishing evidence is the valve response: a healthy valve tracks a 2% step proportionally, whereas this one has a deadband of about 1.5% and then overshoots. That deadband is large relative to the 15% proportional band; the controller must move the output a tenth of its band before the valve responds at all, which sustains a limit cycle.
The 22 s sensor lag is also significant - it adds effective dead time and limits achievable gain - and Td = 0.05 min (3 s) is short compared with that lag, so derivative is probably acting on noise rather than providing useful lead. The 90 s period is consistent with an integral time of 48 s plus valve and sensor lag, which is typical of a stiction-induced cycle. Dominant cause: valve stiction, aggravated by tuning that is too tight for the lag.
2. Describe a test sequence that would confirm the diagnosis and provide data for retuning, without upsetting production.
Read the worked answer
A staged, low-risk sequence:
- Record the current loop trend: controller output, valve stem position, process temperature, and steam flow on the same time axis. Stiction shows a sawtooth in output and a staircase in stem position.
- With the loop in manual, make small output steps (0.5%, 1%, 2%, then down again) and plot stem position against command. Measure deadband, hysteresis, and repeatability in both directions. Do this at several openings, because stiction often varies with stem position.
- Inspect the valve and positioner: air supply pressure, actuator thrust, packing friction, linkage tightness, and positioner zero/span. A positioner can mask or worsen stiction depending on its deadband.
- Verify the temperature sensor: a step test or comparison against a portable reference gives the actual thermowell time constant. The stated 22 s should be confirmed because it sets the achievable controller gain.
- Perform an open-loop bump test at the normal operating point: step the controller output by a safe amount (for example 3–5%) and record the temperature response. Fit gain, time constant, and dead time. This gives the data needed for tuning and shows whether the valve is oversized or undersized.
- Only after the valve is repaired should the controller be retuned; otherwise the new tuning will be compensating for a mechanical fault.
Throughout, keep the process within its temperature limits and have an operator available to take manual control.
3. Recommend remedial actions and explain how the planned 15% duty increase affects the valve and tuning choices.
Read the worked answer
The valve should be repaired or replaced before tuning. Likely fixes: replace or upgrade the positioner, increase actuator thrust or supply pressure, reduce packing friction, and check the stem for binding. If the valve is a plain linear globe with a large deadband, a characterized trim or a rotary control valve with low friction may be more appropriate. The objective is a deadband well below about 0.5% of stroke.
Tuning changes after repair: reduce proportional gain (increase PB), lengthen Ti to reduce the cycle period, and either remove derivative or apply it to measurement with a filter. With a 22 s sensor lag, derivative on error mostly amplifies noise. Consider a small amount of integral-only or PI control first, then add derivative only if the response data justify it. Retune from the bump-test model rather than by trial and error on the running loop.
The 15% duty increase matters in two ways. First, check valve capacity: if the existing 50 mm valve is already near 80–90% open at full load, it will be undersized after the increase, and a saturated actuator will reintroduce limit cycling. Size the replacement for the new maximum flow with margin, and check the steam pressure drop at that flow. Second, the process gain will change with load; a controller tuned at the old duty may be too aggressive at the new one. Retune at a representative high load, and consider gain scheduling or feedforward from water flow if the load varies widely. Instrument the valve stem position so future stiction is visible in the trend rather than discovered as a temperature oscillation.